Also interesting is politics; here the confound is simply that the 19th-20th centuries have seen widespread partisan shifts in particular directions, which means age will correlate strongly with politics unless people are completely spineless.

I strongly suspect that this effect utterly swamps any other effect. Although I'm less confident of this assertion than I was before looking at the study you cited.

Even if attitudes move towards mainstream among older cohorts faster than among younger cohorts, I get the impression that the mainstream is moving faster than the attitude change. A difference between first and second derivatives of attitude? Or am I still relying on stereotype?